Concurrent coevolution of intra-organismal cheaters and resisters
Data deposited at Dryad: doi:10.5061/dryad.80t07
Abstract
The evolution of multicellularity is a major transition that is not yet fully understood. Specifically, we do not know whether there are any mechanisms by which multicellularity can be maintained without a single-cell bottleneck or other relatedness-enhancing mechanisms. Under low relatedness, cheaters can evolve that benefit from the altruistic behaviour of others without themselves sacrificing. If these are obligate cheaters, incapable of cooperating, their spread can lead to the demise of multicellularity. One possibility, however, is that cooperators can evolve resistance to cheaters. We tested this idea in a facultatively multicellular social amoeba, Dictyostelium discoideum. This amoeba usually exists as a single cell but, when stressed, thousands of cells aggregate to form a multicellular organism in which some of the cells sacrifice for the good of others. We used lineages that had undergone experimental evolution at very low relatedness, during which time obligate cheaters evolved. Unlike earlier experiments, which found resistance to cheaters that were prevented from evolving, we competed cheaters and noncheaters that evolved together, and cheaters with their ancestors. We found that noncheaters can evolve resistance to cheating before cheating sweeps through the population and multicellularity is lost. Our results provide insight into cheater–resister coevolutionary dynamics, in turn providing experimental evidence for the maintenance of at least a simple form of multicellularity by means other than high relatedness.
Introduction
Multicellularity
Perhaps the most interesting moments in the history of life are the great transformations in the unit of individuality, when what were previously self-sufficient, functioning individuals, become integrated into a collective, no longer capable of replicating independently (Maynard Smith & Szathmáry, 1995). They are interesting in large part because of the questions they raise about conflict. In order for a higher level of biological organization to form, conflict must be controlled at lower levels. Consider, for example, the origin of multicellularity, one of the six widely recognized major transitions (Bourke, 2011). Multicellularity requires anywhere from a few to many millions of cells to sacrifice for the good of only a minority. Why do the cells in our hands, hearts and brains sacrifice their own reproduction so that our gametes can be passed on?
Inclusive fitness provides one answer (Hamilton, 1964a,b). With each generation, the organism passes through a single-celled bottleneck (i.e. the zygote), meaning that all of the cells within the organism are clonally related. Their relatedness is one (r = 1), so the genetic basis for conflict is effectively eliminated. In addition, any cheater mutation that gets a cell into the germline will be limited to one round of cheating, because in the following generation it will be found in a multicellular organism consisting entirely of its clones (Queller, 2000). Thus, inclusive fitness explains how multicellularity can be evolutionarily stable and may also explain the prevalence of single-cell bottlenecks.
However, questions remain. First, there are alternative explanations for the prevalence of single-cell bottlenecks; they might serve to purge deleterious mutations (Grosberg & Strathmann, 1998) or they might be necessary for complex development (Wolpert & Szathmáry, 2002). The existence of single-cell bottlenecks therefore cannot be taken as strong evidence for the importance of conflict reduction. Second, although most examples of multicellularity have their origin in clonality, there are a few exceptions, such as the social amoeba, Dictyostelium discoideum. If multicellularity did not originate in clonality, what level of relatedness would be required among the cells for multicellularity to be stable? The examples provided by social insects show that extremely cooperative entities (colonies) can be stable without clonal relatedness, in part through coercion or policing (Ratnieks & Wenseleers, 2007), a coevolutionary response of other parties to the evolution of cheating. Are there other forces like this at play that could promote and maintain multicellularity? For example, could noncheaters evolve resistance to cheating? Buss (1987) argued that many aspects of multicellular development evolved from such an interplay between cellular cheaters and resisters. Some of these scenarios are implausible under high relatedness (Queller, 2000), although others, such as control of cell division rates, might not be (Michod, 1997). But if relatedness were low in the evolution of multicellularity, such coevolutionary responses to cheaters might be required for multicellularity to be maintained. Unfortunately, these questions are difficult to explore experimentally. Most multicellular organisms obligately pass through a single-cell bottleneck, such that their intra-organismal relatedness cannot be experimentally manipulated. As D. discoideum becomes multicellular by aggregation, it is a great system in which to explore what is mostly the path not taken.
Dictyostelium discoideum
Dictyostelium discoideum is usually a unicellular amoeba that lives in soil and moist leaf litter, feeding on bacteria. However, when starved, the amoebas aggregate, and form a multicellular slug, which migrates to a new location, at which point about 20% of the cells sacrifice any future reproduction to form a dead cellulose-reinforced stalk. The remaining 80% swarm up this stalk, becoming spores, and forming the sorus (the collection of spores at the top of the fruiting body), which contains thousands of spores (Jack et al., 2011). The stalk facilitates the dispersal of spores by animal vectors to a new location (Smith et al., 2014) where they hatch into single-cell amoebas. Spores within natural fruiting bodies have high relatedness (Gilbert et al., 2007). This is probably largely due to the isolation of founder cells, but also partly due to weak kin recognition systems (Benabentos et al., 2009; Hirose et al., 2011; Gilbert et al., 2012). High relatedness is not necessary for aggregation – different clones mix readily in laboratory experiments (Strassmann et al., 2000) – but the high relatedness in the field explains why multicellularity can be stable even though the fruiting bodies form by aggregation (Fortunato et al., 2003; Ostrowski et al., 2008). This system has been enormously useful for empirical social evolution work, particularly with regard to the origin of multicellularity. Its utility stems from a variety of factors, the most important being that, unlike organisms with single-cell bottlenecks, D. discoideum's intra-organismal relatedness can be manipulated. An experimenter can decide which cells aggregate to form a slug, thus changing the degree of relatedness of the aggregating cells. Indeed, an important prediction of the major transitions view of evolution – that multicellularity is stabilized by self-limitation due to high intra-organismal relatedness – has been tested using D. discoideum (Kuzdzal-Fick et al., 2011).
Cheating of multicellularity in Dictyostelium
Kuzdzal-Fick et al. (2011), starting with a single clone, artificially maintained low relatedness in D. discoideum for 31 rounds of vegetative growth, starvation and spore formation. Low relatedness, maintained by starting each new generation with a random mixture of 106 spores, allowed cheaters that appear by mutation to be favoured by selection. These cheaters cheat by increasing their representation in the sorus, while contributing little or nothing to the stalk.
Further, some of these evolved cheaters were obligate cheaters, which cannot produce fruiting bodies on their own. Obligate cheaters, unlike facultative cheaters, do not modulate their cheating based on their partners (Travisano & Velicer, 2004; Ghoul et al., 2014). Here, cheating entails not sacrificing to form the stalk. In mixtures, this works because the other, noncheating clone forms the stalk. But when alone, an obligate refusal to form stalk means that no spores form either (Ennis et al., 2000; Gilbert et al., 2007), so the organism has no fitness. The importance of this distinction is that cooperation can persist in the presence of facultative cheaters, but obligate nonfruiting, if it sweeps through the population, eliminates cooperation. Obligate nonfruiters also have an added experimental value, because the cheaters can be readily identified when plated out clonally, as they fail to fruit, simply forming a small group of cells.
Even though mutation rates to obligate nonfruiting cheaters are known to be low (Hall et al., 2013), cheaters readily rise to high frequencies when intra-organismal low relatedness provides them with fruiting victims to exploit (Kuzdzal-Fick et al., 2011). At high relatedness, such as occurs in natural fruiting bodies, these mutants do poorly (Gilbert et al., 2007). This raises a question about the evolution of multicellularity. If relatedness were not high, would anything prevent cheating from sweeping through the population? Several mechanisms could be involved (Strassmann & Queller, 2011), one of which is that cooperators could evolve resistance to cheating. Khare et al. (2009) demonstrated that resistance to cheating can be selected for when the cheaters are held constant (not evolving). They presented D. discoideum populations with a cheater for four cycles of selection. They showed that the repeated presence of the cheater selected for resistance to cheating. This experiment suggests that it may be possible for cooperators to evolve resistance to cheating. Hollis (2012) tested coevolving populations of Dictyostelium cheaters and noncheaters, and populations of evolving noncheaters against nonevolving cheaters, and only found evidence for the evolution of resistance when the cheaters were not allowed to evolve. Therefore, it has yet to be demonstrated that a coevolutionary response to cheating can evolve before cheating sweeps through population and multicellularity is lost. Ideally, we would like to know if resistance evolves in real populations, with the cheaters and noncheaters coevolving in real time. Our experiments explore this question.
Resistance experiment
Our experiments test whether D. discoideum can evolve resistance to cheating while cheaters are evolving, before the obligate cheating phenotype sweeps through the population. We used lineages from the Kuzdzal-Fick et al. (2011) experiment, which underwent experimental evolution at low relatedness and evolved cheating. Nonfruiting clones – potential obligate cheaters – increased in the experiment and three of four tested against the (fruiting) ancestor were confirmed to be cheaters (Kuzdzal-Fick et al. 2011). We first confirmed that this ability to cheat the ancestor held for much larger numbers of nonfruiting clones. Then, to test whether resistance had also evolved, we tested the nonfruiters against fruiting clones isolated from their own selection lines. If resistance has not evolved, we would expect to find the same proportion of nonfruiters in the sori of both mixtures (evolved nonfruiters with ancestors and evolved nonfruiters with evolved fruiters). If the evolved fruiters have evolved resistance to cheating, we would expect them to be better than the ancestors at keeping the nonfruiter out of the sorus.
Our results showed that resistance to cheating did evolve. Because we worked with clones from populations that evolved from a single clone, this means that resistance evolved after the obligate cheaters emerged but before they swept through the population. This has implications for both our understanding of the evolutionary dynamics of cheating and resistance (e.g. cooperators can evolve resistance to cheating in real time) and our understanding of the evolution of multicellularity (e.g. there are mechanisms other than self-limitation that can stabilize simple multicellularity).
Materials and methods
Amoebas
We used the ancestor and four experimental lines from the Kuzdzal-Fick et al.'s, 2011 study. The experimental lines had gone through thirty-one rounds of fruiting, with each round initiated by spreading a million cells across a new plate, so any new cheater mutation would be well mixed among victims (Kuzdzal-Fick et al. 2011). All four of these experimental lines had a reported frequency of nonfruiting mutants of 50% or more, which we verified with an initial plating of all the lines.
All lines had been frozen as spores in KK2 Buffer (per litre: 2.25 g KH2HPO4, 0.67 g K2HPO4) with 25% glycerol at −80 °C, as described in the Appendix S1 for Kuzdzal-Fick et al. (2011). For all frozen samples, we thawed them gently (at room temperature), diluted quickly, counted spores using a haemocytometer and then plated 50 spores each onto 10-mL SM/5 plates with 200 μL of Klebsiella pneumoniae in KK2 (OD600 1.5). Plated spores or plated amoebas took between 76 and 80 h to form fruiting bodies. After fruiting, we allowed one week before collecting fruiting bodies. Details of culturing can be found in Kuzdzal-Fick et al. (2011).
Population experiment
The purpose of the population experiment was to determine whether or not there was resistance to cheating in the evolved populations. We achieved this by competing populations of putatively cheating evolved nonfruiters against both the ancestor (which is a fruiter) and populations of evolved fruiters from the same line. Here, population is defined as a mixture of 25 separate clonal plaques of each type. If nonfruiters do cheat the ancestor as expected, resistance to cheating will be established if they cheat the evolved fruiters less or not at all. If the nonfruiter cheats both the ancestor and the evolved fruiter equally, we would expect to see the same number of nonfruiters in the sori for both mixtures. This experiment is illustrated diagrammatically in Fig. 1.
In this experiment, for each line (Kuzdzal-Fick lines: 12, 16, 21, and 24), we initially plated spores at low density (50 spores per plate, 10 plates for each line) and allowed them to fruit. For each line, we then picked and mixed 25 evolved nonfruiters and, separately, 25 evolved fruiters. We did this by categorizing colonies (by phenotype) as either fruiters or nonfruiters, and then for each mixture, picking the leading edge of a colony with a loop, adding the cells to a 1.5-mL microcentrifuge tube with 1 mL HL5 liquid medium (35.5 g L−1), sterilizing the loop and repeating 24 times. Each mixture of 25 clones was created this way for all four lines. We also picked and mixed 4–5 ancestor colonies in the same fashion.
The cell mixtures were cultured in 10 mL HL5 with PSV antibiotic for 2–3 days in 10-mL tissue culture plates to allow growth of enough cells to plate at high density. We split the cultures every 24 h to keep the cells from overcrowding the plates. This was achieved by pipetting the solution up and down with an electronic pipette to lift the amoebas from the bottom of the tissue culture plate, transferring 5 mL of the solution to a fresh tissue culture plate and then adding 5 mL of fresh HL5 with PSV antibiotic to each plate.
We then washed the cells three times using HL5 with no antibiotic, so that we could plate them on SM/5 plates with bacteria as a food source without killing the bacteria. We counted the cells and then plated them with K. pneumoniae in the following ratios: 25% evolved nonfruiters with 75% evolved fruiters, and 25% evolved nonfruiters with 75% ancestor. We also plated each culture on its own (evolved nonfruiter alone, evolved fruiter alone, and ancestor alone), as a control to ensure that all the cells grew up on plates successfully. We plated 2 × 105 cells on each plate to allow thorough mixing of the different clones and allowed 76–80 h for the amoebas to fruit, plus 7 days for the fruiting bodies to be more easily harvestable.
Next, we pooled spores from four fruiting bodies from each plate (evolved nonfruiter with ancestor and evolved nonfruiter with evolved fruiter) and plated 1000 spores from each plate onto 20 SM/5 plates with K. pneumoniae (20 plates per experiment mixture, two mixtures per line, four lines, 160 plates). We diluted the spores across so many plates because we wanted to see individual plaques as fruiters or nonfruiters (our measure of noncheaters and cheaters). We allowed 76–80 h for these to fruit and then scored each colony as either a fruiter or a nonfruiter. We counted the total number of fruiters and nonfruiters for each experimental mixture and used this to determine the proportion of nonfruiters. This served as our measure of the degree of cheating, because it measured how many of the nonfruiters ended up as spore rather than stalk cells.
Individual experiment
The purpose of the individual experiment was to confirm the results of the population experiment at the individual level by looking for resistance to cheating in individual evolved fruiter clones (rather than evolved fruiter populations), and to look for variation among individual evolved fruiter clones. If individual evolved fruiter clones have resistance, we would expect fewer nonfruiters to get into the sori in mixtures of evolved nonfruiters and evolved fruiters than in mixtures of evolved nonfruiters and ancestors. However, if not all evolved fruiters have resistance, or if there are different kinds of resistance in the population, we would expect some mixtures of nonfruiters and fruiters to yield the same number of nonfruiters in the sori as in mixtures with ancestors. The individual experiment is depicted diagrammatically in Fig. 2.
In the individual experiment, we used the three lines that most clearly appeared to have evolved resistance in the population experiment: 12, 16 and 21. We plated these lines and the ancestor from the freezer at low density (50 spores on each of five plates for the three lines). We allowed 76–80 h for the amoebas to fruit and then picked two fruiter colonies and two nonfruiter colonies for each line. We cultured each clone separately in HL5+ PSV as described previously (thus, four cultures for each line: nonfruiter 1, nonfruiter 2, fruiter 1 and fruiter 2). We cultured ancestors in the same fashion.
We cultured the clones for three days. On the second day of culture, we split the 10 mL of culture into two plates. After three days of culture, we washed the cells of antibiotic by centrifuging the cultures in 15-mL centrifuge tubes and adding fresh HL5 with no antibiotic three times. We then diluted the cells to 1 × 107 cells mL−1. We plated six experimental mixtures from each line, testing each of the evolved nonfruiters against both the evolved fruiters from its own line and the ancestor: Evolved Nonfruiter 1+ Evolved Fruiter 1, Evolved Nonfruiter 1+ Evolved Fruiter 2, Evolved Nonfruiter 2+ Evolved Fruiter 1, Evolved Nonfruiter 2+ Evolved Fruiter 2, Evolved Nonfruiter 1+ Ancestor, and Evolved Nonfruiter 2+ Ancestor. The mixtures were of a 75 : 25 ratio, with the nonfruiter always making up 75% of the mixture. We also plated each clone on its own as a control to ensure that each clone was still healthy after being in liquid culture (and that there was no fruiter/nonfruiter contamination). We allowed ten days for the fruiting bodies to reach the stage of having harvestable sori.
We then harvested all of the fruiting bodies for each mixture using a loop. We diluted the spores from each mixture to 1000 spores in 4 mL of K. pneumoniae in KK2 (OD600 1.5). For each mixture, we plated 200 μL of solution onto each of 20 plates, with the aim of having roughly 50 spores per plate. We allowed three days for the spores to hatch, develop and reach the fruiting stages, so we could score them as fruiters or nonfruiters.
Statistical analyses
We conducted all statistical analyses using r version 2.15.2 (released October 2012, R Foundation for Statistical Computing, Vienna, Austria). We arcsine-square-root-transformed all proportion data (number of evolved nonfruiter spores out of total spores in the sorus) to compensate for skew created by the 0 and 1 boundaries of proportional data. We analysed the data from the population experiment with a paired t-test (95% confidence level). We analysed results from the individual experiment using a Welch's two-sample t-test for samples of unequal sizes (95% confidence level) and multiple two-tailed binomial tests (95% confidence level).
For the figures, we plotted relative sporulation efficiency against mixture. Relative sporulation efficiency ratio is the ratio of cheater sporulation efficiency to competitor (ancestor or evolved fruiter) sporulation efficiency. Sporulation efficiency is calculated as fraction of spores in the sorus over fraction of cells in the initial mixture (see Buttery et al., 2013).
Results
Population experiment
The population experiment tested for resistance to cheating in noncheaters from experimentally evolved populations that contained cheaters. We did this by testing whether the evolved fruiters did better than the ancestral fruiters against the evolved cheating nonfruiters.
In mixtures with the ancestor, the proportion of nonfruiter increased significantly (two-tailed, one-sample t-test, t3 = 6.75, P = 0.007). The proportion of evolved nonfruiter rose from 0.25 to a mean of 0.660 (95% CI: 0.463 < u < 0.832), with their sporulation efficiency about seven times that of the ancestor (Fig. 3). In contrast, the proportion of evolved nonfruiter did not significantly change in mixtures with the evolved fruiters (two-tailed, one-sample t-test, t3 = 1.48, P = 0.235). The mean proportion of nonfruiter rose from 0.25 to a mean of 0.344 (95% CI: 0.155 < u < 0.563). A paired t-test rejects the null hypothesis that there is no difference in nonfruiter proportion between evolved fruiter mixtures and ancestor mixtures (Fig. 3, t3 = −8.29 P = 0.004), and therefore supports the evolution of resistance.
Individual experiment
The individual experiment tested for resistance to cheating in individual evolved fruiter clones.
The proportion of evolved nonfruiters increased significantly in mixtures with ancestors (two-tailed, one-sample t-test, t5 = 11.04, P < 0.001). The mean proportion of nonfruiters increased from 0.75 to a mean of 0.916 (95% CI: 0.884 < u < 0.943). The proportion of evolved nonfruiters in mixtures with evolved fruiters did not significantly change (two-tailed, one-sample t-test, t11 = −1.09, P = 0.298). The mean proportion of nonfruiters decreased slightly from 0.75 to 0.682 (Fig. 4, 95% CI: 0.533 < u < 0.813). A Welch's two-sample t-test rejects the null hypothesis that there is no difference in mean proportion between mixtures with ancestors and mixtures with evolved fruiters (t12.9 = −4.22, P = 0.001).
The proportion of nonfruiters increased significantly in six of six ancestor mixtures (Fig. 5, P < 0.05, two-tailed binomial tests, 95% confidence level). However, the tests of nonfruiters with evolved fruiters showed much more variation. In two of the evolved fruiter mixtures (16 F1+ NF2, and 21 F1+NF2), the proportion of nonfruiters did not change significantly (P = 0.564 and P = 0.143, respectively, two-tailed binomial tests, 95% confidence level). In four of the evolved fruiter mixtures, the proportion of nonfruiters went up significantly, and in six of the fruiter mixtures, the proportion of nonfruiters significantly decreased (P < 0.05, two-tailed binomial tests, 95% confidence level).
Discussion
Our results add to a growing body of knowledge regarding the nature and dynamics of cheating and resistance. Numerous studies have explored the nature of cheating, both in Dictyostelium (e.g. Ennis et al., 2000; Strassmann et al., 2000; Buttery et al., 2009, 2013) and across a wide variety of other taxa. For example, cleaner fish cheat by biting their host instead of cleaning (e.g. Bshary & Grutter, 2002), cooperatively scavenging Pseudomonas bacteria cheat by not producing costly iron scavenging siderophore molecules (Griffin et al., 2004), Myxococcus bacteria cheat in cooperative spore formation (Velicer et al., 2000), and fork-tailed drongos mimic alarm calls of pied babblers to gain access to food (Ridley et al., 2007; Flower, 2011) (see Ghoul et al., 2014 for a review of cheating). Our experiment follows up the demonstration by Kuzdzal-Fick et al. (2011) that low relatedness leads to the evolution of obligate cheating, and shows that this cheating is widespread, adding to the stack of empirical evidence that relatedness is important for cooperation and prevents the spread of cheaters. However, factors other than relatedness and kin selection can be important in limiting cheating. This must be the case in between-species interactions where relatedness cannot play a role. Thus, client fish will avoid cheating cleaner fish that bite (Pinto et al., 2011) and legumes may shut off resources to nodules that fail to produce nitrogen (Kiers et al., 2003). Even within species, cheating is sometimes controlled by evolutionary responses among the cheated. For example, in social insects, egg laying by subordinates can be controlled either by dominant queens or through policing by other workers (Queller & Strassmann, 1998). Our experiments, which explored the coevolution of cheaters and resisters in evolving D. discoideum lineages, tested whether cheater resistance could play this role in the evolution of multicellularity.
We first confirmed the result of Kuzdzal-Fick et al. (2011) that most nonfruiters are obligate cheaters that cheat their ancestor. In the population experiment, the proportion of nonfruiters in the sorus, on average, increased significantly in ancestor mixtures (Fig. 3). In the individual experiment, the proportion of nonfruiters increased significantly in six of six ancestor mixtures. The latter result brings the total for both studies to nine of ten.
Our main question is whether this evolution of cheating nonfruiters provoked a coevolutionary response among the fruiters. Both the population and individual experiments show that fruiting clones that have evolved in the presence of nonfruiters (evolved fruiters) are resistant to the evolved nonfruiters' cheating. In the population experiment, there was no significant change in nonfruiter proportion in mixtures with the evolved fruiter, and in both experiments, evolved fruiters did significantly better than the ancestor when tested with nonfruiters.
A possible alternative explanation for this result is that the 31 extra rounds of adaptation to the laboratory environment make both evolved fruiters and nonfruiters better than the ancestor. This seems very unlikely for two reasons. First, better adaptation to the growth environment is not expected to enhance cheating; clones that produce more spores on their own do not typically produce more spores per cell in mixtures (Buttery et al., 2009). Second, the clone used should already have been well adapted to the environment. Kuzdzal-Fick et al. (2011) used a clone that is descended from strain NC4 collected in the wild in 1933 (Raper, 1984), so it had been in the laboratory environment for over 75 years where it has undergone extensive evolution (Bloomfield et al., 2008). The experimental evolution took place in SM/5 medium with Klebsiella aerogenes on agar plates (Kuzdzal-Fick et al., 2011), an environment that would have been commonly encountered in those 75 years. Third, to further guard against possible issues of laboratory adaptation, the Kuzdzal-Fick evolution experiment was initiated with a clone taken from a line previously evolved in the exact same experimental evolution conditions for ten rounds of fruiting and about 100 cell generations (Kuzdzal-Fick et al., 2011). The only real novelty in the experimental evolution environment was the presence of cheaters due to low relatedness. Fourth, if the laboratory adaptation hypothesis were true, it would imply the strange result that, leaving aside the obligate cheating trait itself, the fruiters are consistently evolving more rapidly to the laboratory than the cheaters are.
A potential test of laboratory adaptation by the evolved fruiters would involve competing them against the ancestor, but this would require different and less comparable methods because we could not assess frequencies via incidence of nonfruiting. Moreover, even if it did show that the evolved nonfruiters did better against the ancestor, that would at most add a dimension to our understanding of the selection (e.g. see Asfahl et al., 2015 for an example where the adaptation was nonsocial). It would not take away the component we have demonstrated – that evolved fruiters resist cheating of the nonfruiters and that this advantage would have been in play in the nonfruiter containing environments where they evolved. Resistance cannot be said to be a side effect if it played a demonstrable part in the selection.
Another possible explanation, though not really an alternative one, is that kin recognition and segregation evolved during the course of the experimental evolution so that when fruiters and nonfruiters are mixed, they segregate out and the nonfruiters do poorly on their own. However, we did not see evidence of this. The lawns of fruiting bodies from the mixture experiments were healthy and uniform, without defective fruiting bodies. This is not surprising because, although D. discoideum does have some degree of kin recognition (Ostrowski et al., 2008; Benabentos et al., 2009; Hirose et al., 2011), the resulting segregation is generally rather weak (Gilbert et al., 2012). It seems unlikely that our clones would have evolved much stronger segregation in 31 rounds of fruiting than natural clones have evolved over countless generations in the field. However, the issue could bear further investigation, not as an alternative hypothesis, but as one possible mechanism for the evolution of resistance to the nonfruiting cheaters.
In our experiment, resistance evolved before obligate cheating could sweep through the population, breaking down multicellularity. Prior experiments showing the evolution of resistance either could not test this because they artificially kept the cheaters from increasing or evolving (Khare et al., 2009), or failed to find coevolution of resistance, perhaps due to experimental design (Hollis, 2012). In the latter case, the lack of findings could be due to having only 10 generations of coevolution, perhaps because the main interest in that experiment was selection for cheating rather than resistance.
Our result offers a potential mechanism by which facultative multicellularity remains stable. The lineages used were kept under artificially maintained low relatedness, so the experiment demonstrates that in the evolution of facultative multicellularity, if there were no high population-structure to keep relatedness high, there could be another mechanism by which multicellularity is maintained: the evolution of cheater resistance. This might explain why no lineages went extinct in the original experiment, although many were showing lower spore production (Kuzdzal-Fick et al., 2011). We cannot exclude the possibility that they would go extinct given enough time, but the evolution of resisters should at least slow the process. And even if extinction would ultimately occur with near-zero relatedness, a modest degree of relatedness together with resistance evolution might be sufficient to prevent extinction.
The individual experiment also suggests that there may be variation in resistance phenotypes in the population. The proportions of nonfruiters were much more variable in evolved fruiter mixtures than in ancestor mixtures. In two of the mixtures no cheating occurred, in four of the mixtures the evolved fruiter was cheated, and in six of the mixtures the evolved fruiter appeared to cheat the cheater. This raises the possibility that, although the frequency of resistance in the population may be similar across lineages, not all individuals have evolved resistance and there may be different phenotypes for resistance.
These results provoke interesting questions about the nature of resistance. The population experiment supports previous work (Khare et al., 2009; Hollis, 2012) that showed that resisters can be noble, meaning they do not cheat the cheater, they only prevent it from cheating (Khare et al., 2009; Hollis, 2012). Although this was true in our experiments on average, in a number of our individual tests, the evolved fruiter was not only resistant, but also ignobly cheated the nonfruiting obligate cheater.
Some cheater genotypes have already been identified, but little is known of the mechanism by which cheating works (Santorelli et al., 2013), and nothing is known about resister genetics or mechanisms. One question that remains is how frequency affects the dynamics of coevolution. In our cheating tests, we tested only one mixture frequency, but previous work with both nonfruiting and fruiting cheaters suggests that who cheats does not generally change with frequency (Gilbert et al. 2007; Buttery et al., 2009). Another open question is the amount of variation in the different cheater/resister phenotypes and how they interact with each other. The results from the individual experiment suggest that this would be a valuable path to pursue.
These questions are relevant to the broader research programme on coevolutionary arms races. There has already been some evidence of cheater–resister evolutionary arms races (Ghoul et al., 2014). Among many examples, there is an evolutionary arms race between brood parasitic cuckoos and their hosts, in which cuckoos are selected to cheat their hosts through egg mimicry and their hosts are selected to detect the deception (Davies, 2000; Spottiswoode and Stevens 2010; Langmore et al., 2011; Stoddard & Stevens, 2011). Although coevolution was originally conceived for cases like this that concern interactions between species, the concept has long been extended to within-species reciprocal interactions, such as between the sexes (Arnqvist & Rowe, 2002) or between nuclear and cytoplasmic genes (Werren, 1987). In our experiment, in only 31 rounds of experimental evolution, nonfruiting and fruiting types evolved in response to each other. These results suggest that a coevolutionary arms race could occur among cell types in facultatively multicellular organisms. This would be particularly likely for resisters that are ‘ignoble’, and cheat the cheaters. Pursuing this avenue of research will add to our knowledge of cheating, resistance and evolutionary arms races more broadly.
Cancer may provide another example of where cheater–resister evolution is important in the context of multicellularity. Cancerous cells can be considered cheaters at the intra-organismal level (Nunney, 1999; Bourke, 2011; Ghoul et al., 2014). Understanding how noncheaters can resist cheaters, particularly in a noble way that maintains cooperation at the organismal level, is a potentially valuable approach for research on cancerous cheats.
Conclusion
Our findings demonstrate that in Dictyostelium discoideum, noncheaters can evolve resistance to cheaters when both are evolving together, and that they can do so before obligate cheating sweeps through the population and multicellularity is lost. This offers a mechanism by which, even if low relatedness conditions occurred in the evolution of facultative multicellularity, at least a simple form of multicellularity could be stabilized by the evolution of resistance to cheating.
Acknowledgments
We thank Stu West and jeff smith for helpful discussions and feedback. This material is based upon work supported by the National Science Foundation under grant number DEB1146375 and the John Templeton Foundation grant number 43667.
